Road surfaces (roads) on which vehicles drive are marked or installed with various road markings such as white lines, yellow lines and cat's eyes constituting traffic lane marking lines. Conventionally the recognition of white lines and other such traffic lane marking lines has, as taught by Patent Reference 1, been done by subjecting an image photographed or taken by image photographing means such as a CCD camera and an image processing ECU to differentiation and binarization processing to detect edges in the image and subjecting point sequences of the detected edge (white line candidate point sequences) to Hough transformation to extract approximated linear components.
Further, when, for example, a white line marked on the road surface is itself physically worn through or partially chipped away, or the white line is indistinct in the photographed image owing to low contrast such as may occur during night driving, the white line cannot be accurately recognized, and, as set out in Patent Reference 2, a technique has been proposed for, in such a case, forming a window in the photographed image, judging the degree of white line wear-off from the density of the white line candidate points in the window, and when wear-off is found, superimposing/compositing a white line candidate point sequences extracted from a photographed image taken a given time earlier on/with the white line candidate point sequences extracted from the current white line image, determining a straight line approximating the composited white line candidate point sequences, and recognizing it as the white line.    Patent Reference 1: Japanese Patent Publication No. Hei 6(1994)-42261    Patent Reference 2: Japanese Laid-Open Patent Application No. 2001-236506
As mentioned above, in the technique set out in Patent Reference 2, a chipped-away portion or an indistinct portion is covered or complemented by compositing (superimposing) the white line candidate point sequences extracted from the current white line image with a white line candidate point sequences extracted from a photographed image taken a given time earlier to elongate the white line candidate point sequences, determining a straight line approximating the elongated white line candidate point sequences, and recognizing it as the white line. However, since the superimposing is conducted on the edge image after edge detection, the accuracy of traffic lane marking line (white line) recognition is not necessarily satisfactory.
That is, since the size of a traffic lane marking line distant from the subject vehicle in the photographed image is smaller than that of a closer traffic lane marking line, a case arises in which a distant traffic lane marking line can not be recognized as edges. As a result, the recognition accuracy is not necessarily satisfactory, since point sequences corresponding to a distant traffic lane marking line are chipped away and are not elongated in the composited image.
Therefore, an object of this invention is to overcome the aforesaid drawbacks and provide a traffic lane marking line recognition system for vehicle configured such that point sequences corresponding to a distant traffic lane marking line are prevented from being chipped away in a composited image when compositing photographed images, thereby enabling accurate and unerring recognition of the traffic lane marking line.